On 2022-01-10 12:47 p.m., James Read wrote:
I've been doing some preliminary experiments with PACKET_MMAP style
communication.
With apologies for "snipping", and disclaimer that I am not an nginx
developer, only a long term user.
So, MMAP has given you "preliminary" analysis of what your kernel can do
with your hardware. Would you care to share, in a meaningful manner, any
results that you feel are relevant to any tcp processes - perhaps nginx
in particular?
I'm able to max out the available bandwidth using this
technique.
Available bandwidth? Please define. Is this local, or WAN? Are you on a
56k dial-up modem? or do you have multiple fail-over, load-balanced
fibre connectivity? MMAP to the best of my knowledge, never claimed to
be able to simulate live (live in the sense 'externally processed IP')
tcp/http connections, so what "recognized benchmark" did you max out?
Could Nginx be improved in a similar way?
"improved"? From what and to what? Starting point? End-point? Similar to
what "way"?
You write (below) "a large number of small pages to a large number of
clients..." Large number? 10 to what exponential? I've just looked at
an nginx server that has dealt with ~88.3 GB/sec over the last few
minutes, and cpu usage across 32 cores is bumbling along at less that
3%, temperatures barely 3 degrees above ambient, memcached transferring
nothing to swap.
Either you have badly explained what you are looking for, or, heaven
forfend, you're trolling.
Paul.
Tired old sys-admin.
James Read
Regards
Alex
> On Fri, Jan 7, 2022 at 6:33 PM James Read
<jamesread5...@gmail.com <mailto:jamesread5...@gmail.com>> wrote:
>
>
>
> On Fri, Jan 7, 2022 at 11:56 AM Anoop Alias
<anoopalia...@gmail.com <mailto:anoopalia...@gmail.com>> wrote:
>
> This basically depends on your hardware and network speed etc
>
> Nginx is event-driven and does not fork a
separate process for handling new connections which basically makes
it different from Apache httpd
>
>
> Just to be clear Nginx is entirely single threaded?
>
> James Read
>
>
> On Wed, Jan 5, 2022 at 5:48 AM James Read
<jamesread5...@gmail.com <mailto:jamesread5...@gmail.com>> wrote:
>
> Hi,
>
> I have some questions about Nginx performance. How
many concurrent connections can Nginx handle? What throughput can
Nginx achieve when serving a large number of small pages to a large
number of clients (the maximum number supported)? How does Nginx
achieve its performance? Is the epoll event loop all done in a
single thread or are multiple threads used to split the work of
serving so many different clients?
>
> thanks in advance
> James Read
_______________________________________________
nginx mailing list
nginx@nginx.org <mailto:nginx@nginx.org>
http://mailman.nginx.org/mailman/listinfo/nginx
<http://mailman.nginx.org/mailman/listinfo/nginx>
_______________________________________________
nginx mailing list
nginx@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx
\\\||//
(@ @)
ooO_(_)_Ooo__________________________________
|______|_____|_____|_____|_____|_____|_____|_____|
|___|____|_____|_____|_____|_____|_____|_____|____|
|_____|_____| mailto:p...@stormy.ca _|____|____|
_______________________________________________
nginx mailing list
nginx@nginx.org
http://mailman.nginx.org/mailman/listinfo/nginx
